Predicting 30-day readmissions from skilled nursing facilities using interoperable FHIR data and explainable machine learning
Predicting 30-day readmissions from skilled nursing facilities using interoperable FHIR
Summary
This study proposes a method to predict the risk of 30-day hospital readmission in Skilled Nursing Facilities (SNF) by utilizing interoperable Fast Healthcare Interoperability Resources (FHIR) data and an explainable machine learning model. The analysis was conducted using de-identified FHIR-mapped data from 14,250 SNF residents, incorporating medication, vital signs, functional status, and social risk indicators.
Details
The study addresses the significant challenge of hospital readmissions in Skilled Nursing Facilities (SNF), which impacts healthcare quality, patient safety, and cost management. A key gap identified is that many SNFs lack sophisticated analytical software to integrate clinical and social data for identifying high-risk residents. To address this, the authors developed an interpretable machine learning model based on interoperable Fast Healthcare Interoperability Resources (FHIR) data. The analysis utilized de-identified FHIR-mapped data from 14,250 SNF residents, covering medication, vital signs, functional status, prior utilization, and social risk indicators. A Gradient Boosted Machine (GBM) model was constructed and compared to logistic regression, with performance assessed using metrics like AUROC and AUPRC. Crucially, the study employed SHapley Additive exPlanations (SHAP) for explainability, providing clear, clinically significant explanations of individual risk factors. The findings demonstrate that combining interoperable FHIR data with explainable machine learning enables SNFs to predict readmission risks ethically, transparently, and effectively. This strategy supports policy compliance, privacy, and quality improvement objectives, offering practical value to clinicians, administrators, and policymakers aiming to minimize preventable hospital readmissions.
Technology Note
FHIR（Fast Healthcare Interoperability Resources）は医療データ交換の国際標準。このエントリの関連技術: FHIR
Original content copyright by respective publishers